Smallest.ai привлекла $13 млн на голосовой ИИ, неотличимый от человека

Smallest.ai announced on July 31, 2026, the raising of $13 million in investments to create a new class of voice AI models — systems that make automated phone calls virtually indistinguishable from conversations with a real human.

What Smallest.ai is building and how it works

Smallest.ai develops specialized voice AI models for automated calls. The key declared goal is passing the Turing Test in the voice channel: the person on the other end of the line should not be able to tell they are speaking with an AI agent rather than a live operator. This requires not just a realistic voice timbre, but also the agent's ability to conduct a meaningful dialogue in real time without noticeable delays.

The startup positions its solution as "ultra-fast voice AI that sounds truly human" — this is how TechCrunch describes the company's mission. The round's investor composition and company valuation were not publicly disclosed; the only known figure is the $13 million raised.

Key project parameters:

• Round size: $13 million
• Goal: voice AI agents that pass the Turing Test in real calls
• Technical priorities: ultra-low response latency + natural intonation
• Application areas: phone sales, customer support, service bookings

Why latency is the main barrier for voice AI

For voice AI agents, response latency is often a more critical parameter than the quality of speech synthesis itself. In a live conversation, a person perceives pauses longer than 200–300 milliseconds as unnatural. It is precisely the pauses between responses — not the voice timbre — that most often give away an AI agent in a phone conversation.

Smallest.ai is betting that the voice AI market systematically underestimates latency. The company builds its model architecture around speed as the primary constraint. A perfectly sounding voice with a one-second delay will still fail the Turing Test: the other person will feel the unnatural pauses and realize they are talking to a bot before they even have a chance to evaluate the timbre or intonation.

Key players in the voice AI market include ElevenLabs, Hume AI, PlayHT, and Livekit Agents. Most of them position themselves primarily through the quality, multilingualism, and emotional expressiveness of synthesized speech. Smallest.ai occupies the niche of ultra-fast solutions focused on real calls, where minimal latency matters more than studio recording quality.

Why businesses need AI calls indistinguishable from live ones

Automated calls that pass the Turing Test open up several large commercial directions. First, there is the outsourced call center market, where millions of human operators work today. Second — phone sales and lead generation, where dialogue quality directly affects conversion rates. Third, service communications: service bookings, reminders, delivery notifications.

The key limitation of current voice AI solutions is pauses and mechanical intonation that destroy the illusion of a live conversation and reduce the listener's trust. If Smallest.ai overcomes this barrier, AI agents could replace a significant share of operators in routine scenarios without losing the quality of customer interaction.

"The startup is building voice models designed to make AI calls capable of passing the Turing Test,"

TechCrunch describes the company's mission.

What this means

A $13 million round is a relatively modest sum by the standards of major AI labs in 2026, but typical for an early-stage niche startup with a clear technological specialization. The company's bet on ultra-low latency as its primary differentiating property is an interesting positioning choice in a market where most competitors compete on voice beauty rather than speed.

If the team manages to deliver the declared quality, Smallest.ai has a chance to become an infrastructure provider for a new generation of voice agents — ones that will finally pass the Turing Test not in laboratory conditions, but in real commercial calls.